Machine learning is useles-

No, it is not currently possible for a neural network to simulate Asperger's.

The only useless thing I see here is you and your database.

Same way our brain does it? Pretty obvious if you think about it.

A lot since, it'll save me around 3 seconds every time and the way my life is going I'll be browsing the chans 20/years from now.

Why not?

>Same way our brain does it? Pretty obvious if you think about it.
I...

im using ml to detect frogs

How?

convolutional neural network

holy shit, I've done something similar
>be me, have an unsorted folder with >10k images saved from the chans
>want to post the perfect image in a post, have a vague picture of what the image looks like
>can remember it was red on the left
>decide to use MATLAB's machine learning toolboxes to sort images by colors and color "blob" coordinates
>manually create a training and testing data set of 500 images each for supervised learning (maximum autismo)
>conduct supervised learning on random forest and neural net for image classification
>testing dataset: 97% accuracy
>run code on entire meme images folder
>now have multiple subfolders that organize by categories: mostly COLOR, COLOR on left/right/top/etc, COLOR1 on left COLOR2 on right, etc etc etc
>allow duplicates for easier finding
i wonder if i should rewrite in python/cpp and make a github to build a portfolio. I've got some other random QOL tools i've made for myself through the years, too.

Based
Post the script nigga

its not done

Post it

there's a black man mixed in there

if u aren't a machine posting this then its useless indeed

yeah i know

no. I want to take credit for it. probably going to translate it to python and cpp and throw it up on github. then i'll make a thread here. i'm busy atm so maybe 1-2 weeks from not i'll get around to it

>probably going to translate it to python and cpp
and this will probably be more difficult than just translating, as i'm sure that the machine learning libraries in python and cpp use different algorithms than MATLAB's proprietary ones.

He means where are you collecting them. Are you scraping pictures from Jow Forums and classifying the pictures as pepe and not-pepe.

Why would you not get credit if you posted it?

>Using a CNN instead of simple histogram distance

Never going to make it

machine learning in a nutshell
>we only killed ONE guy with our self driving car for every 300 good trips, what's the problem? It gets smarter every time!

someone can c/p it into their github or something.

its looking at all the thumbnails on Jow Forums, downloading if 90% sure they are frog

can it handle frog color edits?

Attached: roasted.jpg (604x570, 21K)

not those that are that abstract

Retard.
Just go on knowyourmeme and save images from there.
Or better yet, don't frogpost like a fucking faggot on anno domini MMXIX.

I bet you don't have my OC's in there.

>creating oc

If it can be used for downloading frogs, it can be used to filter them out, it's a win-win situation.

I think not those fancy machine learning models are easy to fool

i'd be nice to be able to put all the frogs into 1 concentration folder

if (funny) {
move
}

yes

>Same way our brain does it
What way would that be, exactly?

>20/years
How many years?

If it kills less than humans it arguably is better, but are we willing to accept AI killing 1 in 2000 compared to drunk drivers killing 1 in 1000?

Can you create a peep database? Would save me making ones that already exist.

>but are we willing to accept AI killing 1 in 2000 compared to drunk drivers killing 1 in 1000?
If AI lowers the amount of drunk drivers on the roads then gladly.

Can you dump your pepes, I have a need for them in my GAN.

He's using machine learning to filter frogposts from retards like you

Holy shit do I have to spell it out for you idiots

Has anyone made a pepe gan yet? Seems like fertile ground.

Attached: 1553469895087.png (427x576, 310K)

>not usimg this to create new pepes
Imagine ThisPersonDoesNotExist but with pepe.

what about this?

Attached: Screen Shot 2019-04-11 at 1.22.01 AM.png (2520x1446, 1.57M)

It's a legitimate question, there could be dozens of things he could to
-sort his image collection
-automatically download new frogs
-filter posts
-learn a model that automatically creates new frogs
etc.

I need your folder thanks

Imagine creating a fucking AI solely for singling out le funny frog images. Cringe

Now use a GAN to make new frogs

>le cringe
You're right on the frogs, but drop the faggot buzzwords from Jow Forums and call out other posters so we can kill it

>check if the image is downloaded from reddit
Pretty easy desu

I'd kiss those lips

Holy shit do it.

>loonix
fucking retard

OP is trying to make the perfect frogpost

pajeet

I cant imagine being so braindead to want to do anything with frogs other than filter them, sorry.

do you need to know the source code of Jow Forums pages? e.g. how they name their html elements that contain images etc. ?

are you op

Oh no

just stop browsing Jow Forums(nel) and get a life?

have you considered searching google for frogs

>he doesnt know

>>not usimg this to create new pepes
>Imagine ThisPersonDoesNotExist but with pepe.

OP upload your database and I'll set up a ThisPepeDoesNotExist on github

I made a pol bot and trained it with 4 MB of pol shitpost, was pretty fun and seem to work, but I want to get a non shit tier PC so I can make a Tay bot clone.

How did u create it?

When your finished OP, consider allowing it to output md5 sums so that frogposters can be filtered with 4chanX.

Asking help on the tensorflow channel and two gays of watching youtube lectures.

Not really hard to make, since every model already has a fucking premade library for it.

github.com/gunthercox/ChatterBot

This one is retarded as fuck as to make, just feed him data.

There's a 4plebs datadump on archive.org.

OP, please, I beg of you, create thisfrogdoesnotexist

I want to see randomly generated pepes and some of the eldritch abominations that may result. I'd also love to see some elaborate pepe deepfakes

hydrus network

It's probably quicker and easier to build a time machine and convince Matt Furie's parents to wear a rubber.

Attached: oy.png (395x477, 24K)

Attached: Frog Posting - Pepe.webm (640x360, 1.8M)

share please

underrated lmao

>preggo porn
The patricians porn

How do you do web scrapping on Jow Forums with Python? There are some blockers when I try to use the requests lib

if you could upload a .zip of that folder somewhere for us to download that'd be great

Attached: 1554406056161.png (680x680, 188K)

I've never heard anyone call machine learning useless

>Same way our brain does it?
By the amount of "based and redpilled" replies it has?

>OP upload your database and I'll set up a ThisPepeDoesNotExist on github

Attached: why-arent-we-95qhh3.jpg (500x335, 28K)

Do this nigger.

genius, you could send frogs right into /dev/null where they fucking belong

you dumb retard you're supposed to organize as you go. Now do it manually you cro magnon faggot.

cool, now make a program/extension that autohides and filters every post that contains a pepe image
or send the network and data to me and I'll gladly make it

>mostly COLOR, COLOR on left/right/top/etc, COLOR1 on left COLOR2 on right
I did that without machine learning. It worked all right.

I see nothing of value being done here with ML. It sure has its uses but this is a waste of computing.

If the OP uploads his data it wouldn't be too hard. Nvidia released the source code on that particular NN and it's very easy to retrain on different datasets (in fact in the paper itself they show how it can also be used to generate cats, cars and bedrooms). So yeah. Not hard at all.

Op plz release the pepes

how would it work when there's an image with both wojakt and pepe? do you copy them both to pepe and wojak folder?

do it OP

DONT DO IT OP. THE WORLD DOESNT NEED WHAT YOU'RE DOING

based and redpilled

>filter frogposts from retards like you
pretty sure he's collecting them senpai

I'll get started on the logo